Effectiveness of Containment Measures Against COVID-19 in Singapore: Implications for Other National Containment Efforts

Abstract

Background: We hypothesize that comprehensive surveillance of COVID-19 in Singapore has facilitated early case detection and prompt contact tracing and, with community-based measures, contained spread. We assessed the effectiveness of containment measures by estimating transmissibility (effective reproduction number, (Equation is included in full-text article.)) over the course of the outbreak.

Methods: We used a Bayesian data augmentation framework to allocate infectors to infectees with no known infectors and determine serial interval distribution parameters via Markov chain Monte Carlo sampling. We fitted a smoothing spline to the number of secondary cases generated by each infector by respective onset dates to estimate (Equation is included in full-text article.)and evaluated increase in mean number of secondary cases per individual for each day's delay in starting isolation or quarantine.

Results: As of April 1, 2020, 1000 COVID-19 cases were reported in Singapore. We estimated a mean serial interval of 4.6 days [95% credible interval (CI) = 4.2, 5.1] with a SD of 3.5 days (95% CI = 3.1, 4.0). The posterior mean (Equation is included in full-text article.)was below one for most of the time, peaking at 1.1 (95% CI = 1.0, 1.3) on week 9 of 2020 due to a spreading event in one of the clusters. Eight hundred twenty-seven (82.7%) of cases infected less than one person on average. Over an interval of 7 days, the incremental mean number of cases generated per individual for each day's delay in starting isolation or quarantine was 0.03 cases (95% CI = 0.02, 0.05).

Conclusions: We estimate that robust surveillance, active case detection, prompt contact tracing, and quarantine of close contacts kept (Equation is included in full-text article.)below one.

FIGURE 1.

Cases by infectee category (A) cumulative cases over time, (B) cases by onset date. (Dark blue) Infectee with one or more linked potential infectors who developed symptoms before the infectee and before isolation; (Pink) infectee with travel history to a country with ongoing COVID-19 outbreak in the preceding 14 days or reported frequent interactions with travellers from China in Singapore, but with no epidemiologic links to other cases in Singapore; (Purple) infectee with no known infector but who were associated to a known cluster of cases; (Turquoise) infectee with no known infector and who were not associated to any cluster of cases.

FIGURE 2.

Inferred clusters of three or more cases. Nodes are coloured and sized by the duration at-large from symptom onset to isolation or quarantine. Vertices connect each case to his or her maximum a posteriori source of infection.

FIGURE 3.

Serial interval (A) probability density and (B) cumulative density.

FIGURE 4.

Effective reproduction number over time, R_t. Posterior mean number of secondary cases generated by each case (dot) is plotted against the symptom onset date in the respective case. Posterior estimates of R_t (line) is derived via spline interpolation from the dots in each iteration.

FIGURE 5.

Secondary cases generated by time from onset to isolation/quarantine. For cases with the same duration from symptoms onset to isolation or quarantine, the posterior mean number of secondary cases (dot) and the 95% credible interval (vertical line) is plotted. Linear regression of the mean secondary cases against time from onset to isolation/quarantine is performed at each iteration to derive the posterior incremental number of cases for each day of delay from onset to isolation (turquoise inclined line).

FIGURE 6.

Time from onset to isolation/quarantine over the course of the outbreak. Duration from onset to isolation/quarantine for each case (dot) is plotted against the respective case’s symptom onset date. Linear regression (turquoise line) of the data points shows the trend over the course of the outbreak.